Skip to content

Security

What your security review will ask, answered.

Named accounts, an audit trail you cannot edit, a published subprocessor list, and your own OpenAI key if you want one.

Government review

Built for government security review.

County and state security reviews usually ask the same questions: who can sign in, what they can change, what was logged, which vendors see customer data, how incidents are reported, and how AI is governed. Hazzard was designed around that checklist from the start: named accounts, audit trails, a published subprocessor list, and a compliance program we operate continuously, not a packet assembled after launch.

  • Access and accountability

    Role-based permissions for each organization. Material changes are logged. Hazzard operators use multi-factor authentication.

  • Vendor transparency

    Subprocessors that handle customer data are listed on this page with purpose and the authorizations we have on file.

  • Incident reporting

    A monitored security contact for responsible disclosure (see Report a security issue below).

  • AI governance

    Model use is visible in the workspace. During the pilot, AI can run on Hazzard's managed OpenAI access, or your admin can add your organization's own OpenAI API key so every call bills under your agreement.

Where we are

  • Control frameworkin place
  • GovRAMP membershipin progress
  • GovRAMP verificationplanned

GovRAMP Core verification is planned, not complete. The status in the header and the milestone table below reflect what Hazzard has recorded in its compliance system. We do not claim authorization we have not earned.

GovRAMP

Where we are on GovRAMP.

  1. GovRAMP membershipNot scheduled
  2. GovRAMP Core evidence package downloadedNot scheduled
  3. 90 days of continuous-monitoring evidenceNot scheduled
  4. GovRAMP Core service request submittedNot scheduled
  5. Core verified — listed on the Authorized Product ListNot scheduled
Subprocessors

Who handles customer data.

OpenAI

Core LLM subprocessor for model inference on Hazzard's managed path, in line with GovRAMP-approved AI vendors. Synchronous calls send store:false on OpenAI's Responses API so prompts and completions are not kept for training.

When your organization adds its own OpenAI API key, model calls use that key and bill under your agreement with OpenAI rather than Hazzard's managed access.

Background deep-research is the one managed-path flow that stores an OpenAI response, only so an interrupted run can resume.

Receives: Document text, Page images, Text chunks, Search queries

Disclosure

Report a security issue.

Email chris@hazzard.ai.

How Hazzard uses AI models is described on How we use AI. This page is also available as JSON.